Mimulus Guttatus) Floral Microbiome

Home , Erythranthe, Erythranthe guttata

bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

Floral organs act as environmental filters and interact with pollinators to structure the yellow monkeyflower (Mimulus guttatus) floral microbiome.

María Rebolleda Gómez1 and Tia-Lynn Ashman

1Corresponding author: [email protected]

Affiliation: Department of Biological Sciences, University of Pittsburgh, Pittsburgh PA 15260

Running title: Floral organ shapes monkeyflower microbiome

Abstract

Assembly of microbial communities is the result of neutral and selective processes. However, the

relative importance of these processes is still debated. Microbial communities of flowers, in

particular, have gained recent attention because of their potential impact to plant fitness and

plant-pollinator interactions. However, the role of selection and dispersal in the assembly of

these communities remains poorly understood. We evaluated the role of pollinator-mediated

dispersal on the contribution of neutral and selective processes in the assembly of floral

microbiomes of the yellow monkeyflower (Mimulus guttatus). We sampled floral organs from

flowers in the presence and absence of pollinators within five different serpentine seeps in CA

and obtained 16S amplicon data on the epiphytic bacterial communities. Consistent with strong

micro-environment selection within flowers we observed significant differences in community

composition across floral organs and only a small effect of geographic distance. Pollinator

exposure affected the contribution of environmental selection and depended on the rate and

“intimacy” of interactions with flower visitors. This study provides evidence of the importance

of dispersal and within-flower heterogeneity in shaping epiphytic bacterial communities of

flowers, and highlights the complex interplay between pollinator behavior, environmental

selection and additional abiotic factors in shaping the epiphytic bacterial communities of flowers.

Keywords: Microbiome; Pollination; Dispersal; Community assembly; Serpentine seeps;

Metacommunity; Mimulus guttatus; Anthosphere.

1 Introduction

2 Community assembly is the product of neutral and selective processes (Nemergut et al. 2013;

3 Vellend 2016). In particular, the composition of a community can change through speciation,

4 dispersal, ecological drift (or sampling of individuals and species over time), and environmental

5 selection (Vellend 2016). Environmental selection is a deterministic process and depends on

6 fitness differences between populations (Chesson 2000; Chase & Leibold 2003; Vellend 2016).

7 Neutral processes, in contrast, are independent of niche differences between species and are

8 predicted to be driven by stochastic differences in birth and death (Vellend 2016). Neutral

9 processes can lead to rapid differentiation of communities when dispersal between them is low

10 (McArthur & Wilson 1963; Hubell 2001; Economo & Keitt 2008). Dispersal, in turn, can be

11 deterministic or stochastic depending on species differences in dispersal abilities (Nemergut et

12 al. 2013; Lowe & McPeek, 2014; Evans et al. 2016). The relative importance of neutral and

13 selective processes in community assembly is still subject of much debate (e.g., Hubell 2001;

14 Tilman 2004; Leibold & McPeek 2006; Morrison-Whittle & Goodard 2015) and the contribution

15 of dispersal to these processes can be hard to measure in the field (Evans et al. 2016).

16 Understanding the relative contributions of these processes in host-associated microbiomes is an

17 important first step to understanding the consequences of microbe communities for the host

18 (Costello et al. 2012).

19 Studies of host associated microbiomes have highlighted the importance of selection by

20 the host in shaping its associated microbial communities (e.g., Rawls et al. 2006; Ofek-Lalzar et

21 al. 2014; Pratte et al. 2018). The host can favor colonization and growth of certain microbes over

22 other through diverse mechanisms like: immune system activity (e.g., Donaldson et al. 2018),

23 host secretions (e.g., Schluter & Foster 2012; Ofek-Lalzar et al. 2014), or specific environmental

1 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

24 characteristics like high osmolarity in flower nectar (Herrera et al. 2010). These effects,

25 however, can be overpowered by dispersal from other hosts or the host environment (Burns et al.

26 2017). Thus, understanding the relative contributions of drift (i.e., neutral) and selective

27 processes in the host can provide insight on the drivers of host-associated microbiome assembly,

28 their changes over time ( e.g. Burns et al. 2016), and the potential sources of these microbes (e.g.

29 Venkatamaran et al. 2015). Recently, flower-associated microbiomes have been established as an

30 excellent system to study community assembly and metacommunity dynamics (Belisle et al.

31 2012; Shade et al. 2013, Vannette & Fukami 2017; Chappell & Fukami 2018).

32 Flowers are multi-purpose reproductive structures and microbial communities of flowers

33 can have a large impact on plant fitness by directly affecting the survival and reproduction of the

34 plant (e.g., Alexander & Antonovics 1995), or through effects on pollination (Vannette et al.

35 2013; Herrera et al. 2013; Schaeffer et al. 2014; Schaeffer et al. 2017; Rering et al. 2017).

36 Understanding microbial community assembly in flowers, can highlight important, and

37 underappreciated, ecological processes affecting floral evolution and plant-pollinator

38 interactions.

39 Despite significant variation across floral organs, and potential effects of microbes of

40 anthers, pollen, styles and stigma on direct fitness effects (not mediated by pollinators), most

41 studies of microbial communities associated with flowers have concerned microbes of nectar

42 (e.g., Herrera et al. 2010; Belisle et al. 2012; Pozo et al. 2016; Mittelbach et al. 2016; Vannette

43 and Fukami 2017). These studies have shown the importance of pollinators in shaping some of

44 the assembly patterns of these microbiomes (Belisle et al. 2012; Herrera et al. 2013; Vannette

45 and Fukami 2017). But there is substantial variation in microbial composition that is not

46 explained by the presence/absence of pollinators (e.g. Vannette and Fukami 2017), and the

2 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

47 source of most floral microbes remains unknown. Each floral organ is likely to create unique

48 conditions for the establishment of bacteria (Aleklett et al. 2014; Junker & Keller 2015).

49 Pollinators that transport microbes have different behaviors on flowers and could create varying

50 opportunities for contact with floral structures (Laverty & Plowright 1988; Russell et al. 2019).

51 Yet we have a poor understanding of the extent to which floral organs and their interaction with

52 pollinators creates heterogeneity in microbial communities within flowers.

53 In this paper we address the relative importance of neutral (i.e., drift and passive

54 dispersal) and selective processes (i.e., organs that create unique habitats within the flower), as

55 well as their interaction with pollinator-mediated dispersal in shaping epiphytic bacterial

56 communities in a flower with no nectar production. If neutral effects are the main factor

57 explaining community assembly, then we expect that: different organs will have a random

58 phylogenetic representation of the whole flower metacommunity and that the most abundant

59 microbes in the whole metacommuity will also be the most frequent in the different organ

60 samples. In addition, if communities geographically farther apart are less likely to share

61 microbial migrants, then, under a neutral model, we expect that these distant communities will be

62 more likely to diverge as a result of ecological drift (Hubell 2001; Soininen et al. 2007). Thus,

63 we would expect that spatial location of the plant in the habitat and not floral organ to explain

64 most of the differences between communities, and that community differentiation (beta diversity)

65 will increase with geographical distance. In contrast, if the different floral organs act as selective

66 microenvironments, then we expect that floral organ and not plant geographic location will be

67 the main determinant of community composition. In addition, if pollinators are the major agents

68 of microbial dispersal, then we would expect pollinator exclusion to affect differentiation among

69 locations or organs. Specifically, pollinators could homogenize communities by transporting

3 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

70 microbes across large spatial scales, or they could increase differentiation by moving microbes

71 mainly within local patches. Finally, high rates of pollinator-mediated dispersal could

72 overwhelm the effects of local dynamics of environmental selection within the flower.

73 Materials and Methods

74 Study system and species

75 The yellow monkeyflower, Mimulus guttatus (Erythranthe guttata, Phyrmaceae) is self-

76 compatible, hermaphroditic and insect-pollinated annual/perennial herbaceous plant that is a

77 dominant component of the serpentine seep communities in northern California (Harrison et al.

78 2000; Freestone & Inouye 2006; Arceo-Gómez & Ashman 2011). Flowers are zygomorphic and

79 tubular, produce little or no nectar, and there is high variability among flowers in the quality of

80 the pollen rewards (Robertson et al. 1999; Wu et al 2008). In the field, monkeyflowers interact

81 with a variety of insect pollinators of varying behaviors and sizes (Arceo-Gomez & Ashman

82 2014; Koski et al. 2015). As a result, seep monkeyflower is highly generalized and well-

83 connected within the pollinator networks of these serpentine seeps (Koski et al. 2015).

84 Monkeyflowers were studied within five seep communities at the McLaughlin Natural

85 Reserve in northern California, USA (Table S1). These seeps are characterized by a high

86 diversity of flowering species restricted by abiotic factors (e.g., water availability) and are

87 separated in space by a grassland matrix. Therefore, seeps can act as discrete replicate

88 communities (Harrison et al. 2000; Arceo-Gómez and Ashman 2014).

90 Experimental set up and epiphytic bacterial sampling

91 At the height of flowering in 2017 (May 9-20) we established five transects along five serpentine

92 seeps with three (RH1, RHA and RHB) or four (TP9 and BNS) sampling points (Table S1; Fig.

4 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

93 1A). Within each transect the locations of the sampling points provided a range of inter-seep

94 distances from ~10m to ~100m (Fig. 1B). The geographical position of the longest sampling

95 point within a seep was recorded as GPS coordinates (Table S1). For the shortest distances we

96 used the distances measured in the field. Using qGIS 2.18.10 (qGIS development team, 2016) we

97 projected all coordinates on WGS84/UTM Zone 10N to obtain a distance matrix in meters.

98 Within each seep and at each location, we set up paired control and pollinator-exclusion

99 cages (treatments). Cages were constructed from PVC and tulle (Joann Fabrics, ITEM #

100 1102979). The control cages had open sides to allow for visitation (Fig. 1C). Wearing sterile

101 gloves, we marked the petiole of several flower buds per plant within a cage with a permanent

102 marker. After marked flowers were open for 3-4 days we carefully dissected the organs of three

103 flowers from two to three different plants using sterile forceps. The stamens (anthers and

104 filaments), petals (only the corolla, without the calyx) and styles with stigma (no ovary) were

105 stored in separate sterile vials (Fig. 1D). To obtain enough DNA, for each sampling location-

106 treatment combination we pooled replicate organs from three different flowers, ending up with

107 one sample for each organ, for each treatment and location (102 samples of floral organs).

108 To get a better idea of the potential sources of floral microbes at each location we also

109 sampled a basal aerial leaf from each monkeyflower plant, soil (one random location per seep)

110 and flowers from the co-flowering community (39 samples total). The community samples were

111 a mix of five flowers representing the species in the local community (2 x 3m plot; Table S1).

112 All samples were collected at the same time (within a week of each other) to minimize changes

113 in co-flowering community, pollinator community and other environmental variables like

114 temperature.

115

5 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

116 DNA extraction and 16S rRNA amplicon sequencing

117 To obtain the epiphytic bacterial communities of the flower organs (or leaves or flower

118 communities) we washed samples with 1ml of sterile phosphate buffered saline (PBS) and

119 vortexed them for ten minutes to detach bacterial cells from the tissue (in previous tests we did

120 not observe differences in colony forming units between five minutes of sonication with a small

121 jewelry sonicator and ten minutes of vigorous vortexing; data not shown). We concentrated the

122 microbial cells and used only the bottom 250µl of the pellet for DNA extraction (avoiding the

123 floral tissue). For our soil data we used 200g of soil directly in our DNA extraction protocol.

124 We extracted DNA from all samples using DNeasy PowerSoil Kit (Quiagen). We added a

125 control for DNA extraction and sampling in the field (maintained a tube with 250µl of sterile

126 water opened in the field for ten minutes). Both of our controls failed to amplify. We sequenced

127 the 16S rRNA gene V4 hypervariable region using one run in the Illumina MiSeq platform

128 (Illumina, CA, USA). We used the 515FB-806RB primer pair

129 (FWD:GTGYCAGCMGCCGCGGTAA; REV:GGACTACNVGGGTWTCTAAT) and use

130 paired-end sequencing of 150 base pair per read (Caporaso et al. 2012). All of the sequencing

131 procedures including the library preparation, were performed at the Argonne National

132 Laboratory (Lemont, IL, USA) following the Earth Microbiome Project protocol

133 (http://www.earthmicrobiome.org/protocols-and-standards/16s/) with 12-bp barcodes. To reduce

134 chloroplast contamination, we used peptide nucleic acid (PNA) clamps (Lundberg et al. 2013).

135 All sequences will be made available in NCBI’s Short Read Archive.

136

137 Sequence processing

6 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

138 We used PEAR v0.9.10 paired-end merging (Zhang et al. 2014). After sequencing we obtained a

139 total of 2.1 10ͫreads and we were able to successfully pair 1.9 10ͫ (92%). After merging

140 our reads, we re-assigned the barcodes to the merged reads with a custom script written by

141 Daniel Smith (https://www.dropbox.com/s/hk33ovypzmev938/fastq-barcode.pl?dl=1").

142 Subsequently, we demultiplexed our samples and removed low quality reads (Phred quality

143 scores < 20), aligned them with PyNAST (Caporaso et al. 2010), and assigned taxonomy (OTUs

144 at 97% similarity) with the RDP classifier (Wang et al., 2007), using the Greengenes database

145 (13_8 release), as implemented in QIIME v.1.9.1 (Caporaso et al. 2010b). We removed

146 mitochondria and chloroplast sequences as well as OTUs in low abundance (often spurious)

147 across samples (>0.0005% mean abundance) according to recommendations based on

148 simulations (Bokulich et al. 2012).

149 OTUs are a conservative measure of variation that clusters together sequences within

150 97% similarity. Sequencing errors often fall within that 3% of variation that is allowed within the

151 same OTU and are, thus, not interpreted as meaningful biological variation. However, OTU

152 clustering also losses much of the finer biological variation. DADA2 is a model-based approach

153 for correcting amplicon errors while maintaining sequence level variation (i.e., amplicon single

154 variants or ASVs; Callahan et al., 2016). To evaluate the robustness of our results, we also

155 processed our data following the DADA2 pipeline. After assessing the quality of our reads, we

156 trimmed the last ten base-pairs of our reverse reads but left our forward reads untouched to be

157 able to merge them. Most reads (1.32 10ͫ or 94%) were kept after quality filtering and

158 trimming.

159 After merging reads (we were able to merge 1.2 10ͫ of the filtered and trimmed reads,

160 92%), we removed chimeras and sequences either too short (less than 248bp) or too long (more

7 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

161 than 256bp), and then, we removed chloroplast and mitochondrial sequences. Overall, results

162 from the QIIME 1.9 pipeline (OTUs) and DADA2 (ASVs) were consistent. However, DADA2’s

163 finer resolution tended to amplify stochastic variation between individual samples (see results

164 and discussion). Thus, in this paper we present the results from our OTU (QIIME 1.9 pipeline)

165 data and discuss when both analyses are discordant. Both pipelines are available in gitHub

166 (github.com/mrebolleda/OrganFilters_MimulusMicrobiome).

167

168 Pollination observations

169 To evaluate the effects of pollination intensity on the microbiome assembly, at each sampling

170 location we observed a similar number of flowers within a 2 x 3m plot and recorded floral

171 visitors for fifteen minutes. A visit was recorded as contact with a monkeyflower or a

172 ‘community’ flower in our plot (we did not count as separate visit multiple visits by the same

173 insect to the same flower, and we only counted the first three visits of the same insect, thus we

174 scored visitation at the plot level). We scored visits to focal monkeyflowers as ‘external’ when a

175 visitor contacted petals or ‘internal’ when an insect contacted anthers or stigma within the

176 flower. We classified each visit by the functional group of the insect following Koski et al.

177 (2015). Each sampling location was observed twice between 9:20 and 12:00, and twice between

178 12:00 and 16:00. For each location we calculated the visit rate to focal monkeyflowers and to the

179 community of flowering plants in each plot (visits/plot/hour).

180

181 Analyses of species composition

182 Due to evolutionary divergence in chloroplast sequences (Fitzpatrick et al. 2018), despite using

183 the PNA clamps ~80% of our reads were chloroplast and mitochondria sequences and were

8 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

184 removed from analysis resulting in samples with anywhere from 211 to 179975 reads and a

185 median coverage of 5986 reads per sample (Fig. S1A-B). To facilitate accurate comparisons

186 across samples, we subsampled each to an even sequencing depth of 1200 reads. This number

187 was chosen as a balance between the depth of sampling within each bacterial community and the

188 number of communities (Fig. S1C-D). To minimize the potential effects of stochastic variation

189 due to low coverage, we obtained an average sample from each of our communities after 10,000

190 subsamples with replacement. In addition, we performed all of our analyses with data normalized

191 through a variance-stabilizing transformation as implemented in “DESeq2” (Anders & Huber

192 2010; Love et al. 2014). We obtained the same overall results with both analyses (data not

193 shown), however we present those using our rarefied community for diversity analyses because

194 the results were slightly more conservative. All statistical analyses were performed in R (3.4.4; R

195 Core Team 2017).

196 To characterize the epiphytic microbial communities of the monkeyflowers we calculated

197 four distance matrices using “Sørensen” (only richness), “Bray-Curtis” (relative abundance),

198 “Unweighted Unifrac” (relative abundance and phylogenetic distance) and “Weighted Unifrac”

199 (relative abundance and phylogenetic distance) (Lozupone & Knight 2005). These indices of

200 beta diversity allowed us to evaluate the robustness of our results as well as the importance of

201 relative abundances and phylogenetic information in our flower samples. We performed a

202 PERMANOVA to evaluate the effects of floral organ, pollinator treatment and seep for each of

203 the community distance matrices. We used the full model including as factors: floral organ, seep,

204 pollinator and the two-way interactions. We evaluated assumptions of homogeneity of group

205 dispersion using the function betadisper (Anderson 2006). Only Bray-Curtis showed

206 heterogeneity in the dispersion across organ samples.

9 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

207 In addition, we evaluated the degree of phylogenetic clustering or overdispersion of

208 communities (from a particular floral organ and pollinator treatment) as the deviation from

209 expected phylogenetic diversity. To calculate phylogenetic diversity, we used the total branch

210 length of a given sample (Faith 1992) and the expected phylogenetic diversity was calculated

211 through binomial sampling of the whole metacommunity tree (O’Dwyer et al. 2012). In this case,

212 we defined our metacommunity as all of the monkeyflower samples together. Expected and

213 observed phylogenetic diversities were calculated using the picante package (1.7; Kembel et al.

214 2010).

215 Next, to evaluate the relationship between geographical distance and community

216 differentiation, we calculated bacterial community composition distance matrices using Bray-

217 Curtis and Unifrac for each floral organ within each pollinator treatment. We then performed

218 Mantel tests on these matrices using the “vegan” package (2.4-6; Oksanen et al. 2018). We

219 adjusted p-values to account for multiple testing using false discovery rate correction (Benjamini

220 & Hochberg 1995). We also assessed the degree of concordance between flower community

221 composition and microbial community composition at each location through Procrustes analysis

222 comparing a non-metric multidimensional scaling (NDMS) of data for each pollinator treatment

223 and organ combination.

224 Finally, to evaluate the uniqueness of each community we obtained a “core microbiome”.

225 Comparing overlap in “core” taxa only, provides a way to minimize inflation of non-shared taxa

226 by excluding OTUs that might be present only in a few samples. For this analysis we defined the

227 “core microbiome” as the OTUs that were shared by at least 20% of the samples of a given

228 organ/treatment (the maximum cut-off that still provided a large sample of more than a 100

229 OTUs). In other words, if an OTU was not present in at least 20% of the samples of a given

10 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

230 organ in a given treatment then it was not considered part of the core for that organ and in that

231 treatment. With this list of OTUs we calculated overlap across organs.

232

233 Neutral model fit

234 To determine the potential importance of neutral processes to community assembly, we

235 evaluated the fit of a neutral model for prokaryotic communities (Sloan et al. 2006; Burns et al.,

236 2016). This model is based on the idea that, under neutral assumptions, taxa that are more

237 abundant in the whole metacommunity are more likely to occur in multiple patches (floral

238 samples). Thus, with a single free parameter m (that describes the migration rate) the model

239 predicts the relationship between the frequency to which taxa occur in a series of communities

240 (in this case each organ) and their abundance in the whole metacommunty (all of the

241 monkeyflower samples together).

242 Using the Akaike information criterion, the fit of this neutral model was compared to a

243 null-model (in this case a binomial distribution) that represents random sampling from the “pool

244 community” without drift or dispersal limitations (Sloan et al. 2007; Burns et al. 2016). The 95%

245 confidence intervals around the model were calculated by bootstrapping with 1000 samples.

246 OTUs within the 95% CI were considered to fit neutral expectations of distribution in the

247 metacommunity. OTUs outside the confidence intervals were separated in two categories:

248 overrepresented (present in more samples than would be predicted by their mean relative

249 abundance) and underrepresented (present in less samples than would be predicted by their mean

250 relative abundance). The relative proportions of OTUs in these different categories among floral

251 organs where evaluated with chi-square tests of independence with Bonferroni correction for

252 multiple tests.

11 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

253 To determine the extent to which over- or underrepresented taxa are exclusive to each

254 organ (instead of shared across two or more of the floral structures), we generated a null model

255 to determine, given the number of OTUs in each category, how many would we expect to be

256 shared between at least two organs controlling for the number of OTUs (independently their

257 category). We repeated this sampling process 10,000 times to generate a null distribution for

258 comparison with our observed values. To determine the proportion of OTUs shared across our

259 organ samples we used the get.venn.partitions function in the “VennDiagram” package (1.6.20;

260 Chen & Boutros 2018).

261

262 Analyses of potential sources of monkeyflower microbial communities

263 To understand the relation between flower communities with other local communities of

264 microbes, we obtained the differences in OTU composition (beta diversity) between our focal

265 monkeyflower epiphytic microbial communities (for a given organ) and the communities acting

266 as potential sources (i.e., the rest of the floral organs, monkeyflower leaves, co-flowering

267 community or soil). We calculated total beta diversity and the components due to nestedness and

268 turnover (Baselga 2010) between the focal communities and the potential sources for each seep

269 using the R package “betapart” version 1.5.0 (Baselga & Orme 2012). A large contribution of

270 “nestedness” means that differences between communities are mostly due to subsample (species

271 losses) from the more diverse to the less diverse community. Whereas, a large contribution of

272 “turnover” indicates species replacement across communities (Baselga 2010). For each seep we

273 compared an average source community (to minimize variation due to differences in sample

274 sizes of sources) with all of the communities for a given floral organ in a given seep.

12 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

275 To evaluate whether potential sources differed in their compositional distance to our focal

276 communities, and to investigate the contributions of nestedness and turnover to their overall beta

277 diversity, we performed an ANOVA with beta diversity as the response variable, and component

278 (nestedness or turnover), floral organ and source as factors. We did not include pollinator

279 treatment in this analysis because communities from both of our pollinator treatments showed

280 the same patterns (see results).

281 Beta diversity indicates differences between potential sources and our focal communities,

282 and the decomposition of beta diversity into nestedness and turnover components highlights the

283 ways in which those communities are different. However, to identify the likely sources of our

284 focal communities (and the uncertainty around these calls) we used SourceTracker as

285 implemented in R (version 1.0.1). SourceTracker is a Bayesian approach that models a sink

286 community as a mix of potential sources, allowing for assignment into an unknown source when

287 part of the sink (focal community) is not like any of the sources (Knights et al. 2011). The code

288 for all of the analyses is available in gitHub

289 (github.com/mrebolleda/OrganFilters_MimulusMicrobiome).

290

291 Results

292 Floral organs are the main factor explaining variation epiphytic bacteria community

293 composition

294 PERMANOVA results were fairly consistent across all beta diversity indices: pollination

295 treatment (exclusion/control), seep and floral organ and their two-way interactions explained

296 between 26% to 33% of the total variation in epiphytic bacteria of monkeyflowers (Table 1). In

297 general, our model was slightly better at explaining species composition alone (Sørensen and

13 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

298 Unifrac) than abundance (Bray-Curtis and weighted Unifrac). The presence of OTUs (more than

299 their relative abundances) distinguishes between organs (Table 1; Fig. S2). Floral organ was

300 significant in its contribution to community composition across all the different distance metrics

301 (Fig. 2A, Table 1, Fig. S2), and alone explained between 4% and 11% of the variation (Table 1).

302 Seep was marginally significant (α=0.05) in all comparisons but Bray-Curtis (Table 1) and

303 pollinator treatment was not significant, but the interaction between pollinator treatment and

304 organ was marginally significant across indices (explaining ~3% of the variation; Table 1, Fig.

305 2A). Overall, floral organ was the only factor that was consistently significant, with seep and

306 interactions between factors being marginally significant in half or more of the analyses (Table

307 1).

308 Higher taxonomic resolution (using ASVs instead of OTUs; see methods) provides

309 additional support for organ as the main factor structuring microbial communities in flowers.

310 Using ASVs we might expect an increase contribution of stochastic and local processes (where

311 specific strains might only be locally distributed, or present in only a few samples). Despite this

312 increased stochasticity, organ is still significant across all diversity indices (Table S2). However,

313 seep and pollinator by organ interaction are no longer significant (Table S2).

314 Consistent with organs acting as environmental filters, microbiomes of a given organ in a

315 given treatment tended to have less phylogenetic diversity than expected from random sampling

316 of the whole floral bacterial metacommunity (Fig. 2B). Stamens and styles were dominated by

317 OTUs from the Pseudomonadales order, while the most abundant OTUs associated with petals

318 were more evenly distributed across the Pseudomonadales, Bacterioidales and Clostridiales (Fig.

319 S3).

14 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

320 Despite evidence of flower organs contributing to community differences across

321 microbiome samples, we observed high variation across samples of the same organ (Fig. 2A, Fig.

322 S2, Fig. S3) and an exponential decrease in the number of OTUs shared by an increasing

323 percentage of samples of a given organ (Fig. S4A). Furthermore, a large majority of core

324 microbes were shared across organs, with no significant differences between the proportion of

325 shared vs. unique OTUs between pollinator treatments (Fishers exact test, p=0.225; Fig. 2C).

326 Among core microbes we found some of the most abundant genus in all organs: Acinetobacter,

327 Pseudomonas, Bacteroides and Corynebacterium. In addition, we found some common flower

328 associated genus like Erwinia and Lactobacillus as well as some unidentified Prevotella and

329 Streptococcus (more commonly associated with human hosts). Across both treatments, petals

330 had the largest number of unique OTUs even though we did not observe differences in alpha

331 diversity at our sampling effort (Fig. 2D).

332

333 Floral organs act as selective filters

334 We observed no significant relation between geographical distance at the measured spatial scales

335 (meters to kilometers) and beta diversity for petals, and styles, with or without pollinators.

336 Across treatments, communities displayed the same level of differentiation (beta diversity). Only

337 stamens showed increased community differentiation when exposed to pollinators than in the

338 control, and only the stamens exposed to pollinators showed a significant relation of increasing

339 beta diversity with increasing distance (Mantel test with 999 permutations, r=0.266, P=0.038)

340 (Fig. 3A). This relationship was similar to that seen in the leaf samples (Mantel test with 999

341 permutations, r=0.258, P=0.032; Fig 3A).

15 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

342 Consistent with ecological selection across different floral organs, we found a large

343 proportion of OTUs that are either overrepresented or underrepresented under neutral

344 assumptions (Fig 3B). The neutral model only explains between 20% and 37% of the variation in

345 the distribution of taxa (Fig 3B), but it is a better fit than a model ignoring drift and migration

346 (Table S3). To assess deviations from neutrality across organs, we compared the proportion of

347 OTUs in each category (i.e., overrepresented, underrepresented and neutral) across the three

348 floral organs. We observed significant differences in the proportion of OTUs in each category

ͦ 349 across the three different organs ( ʚͨʛ 19.638, P= 0.0006). There was no difference between

ͦ 350 styles and stamens ( ʚͦʛ 0.551, P= 1), but petals had a higher proportion of overrepresented

ͦ ͦ 351 and underrepresented taxa than styles ( ʚͦʛ 15.344, P= 0.001) or stamens ( ʚͦʛ 11.327,

352 P= 0.01). In general, core taxa (present in at least 25% of samples from a given organ) were also

353 some of the most abundant (although this is not always true in the case of some overrepresented

354 taxa that are present in more than 25% of the samples despite having an overall low frequency,

355 and some of the underrepresented taxa, that are not present in even 25% of the samples, but

356 when they are they might be in high abundance; Fig 3B).

357 The neutral models perform worse when we separate the data by pollinator treatments

358 (i.e. exposed control and pollinator exclusion; Table S3). This reduction could be due to smaller

359 sample sizes. Nevertheless, the neutral models still explain between 15% and 26% in the

360 different treatments of stamens and styles, while these models explain less than 7% of the

361 variation in the distribution of taxa from petals (Table S3). For the most part, we did not observe

362 strong differences in the fit of neutral models when comparing our two pollinator treatments

363 (Table S3, Fig. S4). However, in the presence of pollinators the neutral model explained the data

16 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

364 better than in our pollinator exclusion treatment (R2 Control= 0.15 vs. R2 Exclusion= 0.26 and

365 AIC Control= -79.5 vs. AIC Exclusion= -207.3; Table S3).

366 If each organ is selecting for a particular microbial community, we would expect that

367 over or underrepresented taxa for a particular organ will not be shared across different organs.

368 Whereas OTUs that are distributed according to a neutral model would be distributed more or

369 less randomly across the whole flower. According to our expectations, OTUs distributed

370 according to neutral expectations are shared across two or more organs in the same proportion as

371 we would expect by chance (Fig. 3C). Instead, over and underrepresented OTUs are shared

372 between organs more than we would expect by chance (Fig. 3C). This effect is stronger in the

373 control treatment; in the presence of pollinators over- and underrepresented taxa are more likely

374 to be shared among organs than when pollinators are excluded (Fig. S4).

375

376 Dispersal can overwhelm the effects of organ selection

377 Despite evidence of organ-specific selection on bacterial communities, we also observed a

378 marginally significant interaction between organ and pollinator treatment (although only in the

379 OTU data). In the pollinator exclusion treatment, almost twice as much of the total variation is

380 explained by floral organ, relative to the control (Fig. 4A). This is true even in the ASV data set,

381 where, organ explains more variation in the pollinator exclusion treatment than in the control

382 (except in the case of Bray-Curtis; Figure S5). In addition, floral organs of pollinator excluded

383 plants have a larger proportion of unique OTUs across a variety of cut-off values for the core

384 microbiome (0-60%; Fig. S4B).

385 Pollinator service was heterogeneous in both quality and quantity. Some insects mainly

386 encountered the external parts of M. guttatus, whereas others contacted the internal reproductive

17 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

387 organs of the flower. The community of ‘external’ visitor insects differed significantly in

388 composition from the community of insects visiting the ‘internal’ parts of the flower (Fishers

389 exact test, p>0.0001; Fig 4B). In addition, visitation varied in rate across location with visitation

390 to the yellow monkeyflowers varying from 6-55 mean visits/plot/hour (Table S1). We observed

391 no significant correlations between the total number of visitors (internal and external) and the

392 pairwise beta diversity between petal samples of the pollinator exclusion and the control

393 treatment. In contrast, when we consider only internal visitors (those that might be in direct

394 contact with anthers and stamens) we observed a positive correlation for the stamens (increasing

395 in strength as we account for abundance and phylogenetic information in the beta diversity index

396 used; Fig 4C). This pattern is maintained when looking at the ASV data. In contrast, there are no

397 clear patterns for petals and styles, and all significant correlations (styles using Bray-Curtis and

398 petals using Unifrac; Figure 4C) are lost when analyzing the ASV data (Fig S7).

399 We did not observe a clear association between the co-flowering community composition

400 and microbial community composition for none of the organ/pollinator treatment combinations

401 (Table S4). Alpha diversity was not correlated with pollinator visitation rates, co-flowering

402 community abundance nor co-flowering community diversity (Fig S8).

403

404

405 Potential sources of floral microbes

406 Our results suggest that pollinator-mediated dispersal of microbes can affect community

407 assembly and possibly override the contribution of environmental selection from different floral

408 organs. But the ultimate sources of microbial communities of flowers remain unclear. Using

409 decompositions of betadiversity, we can ask how much organ-specific communities differ from

18 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

410 other microbial communities that could act as sources (i.e., soil, M. guttattus leaves,

411 heterospecific co-flowering neighbors or remaining parts of the M. guttattus flower) and how

412 much of these differences is due to replacement of OTUs (turnover) or loss of OTUs in a nested

413 manner from the potential sources (or more diverse communities). Levels of beta diversity were

414 high across all comparisons (0.77±0.012 SE), indicating differentiation of our focal bacterial

415 communities from all other potential sources. Across floral structures beta diversity was highest

416 when focal organ-specific communities were compared with soil and lowest in comparison with

417 the neighboring heterospecific flowering community. Surprisingly, these patterns of

418 differentiation from potential sources, were not different across pollinator treatments (Fig. 5).

419 The contributions of nestedness and turnover were significantly different (ANOVA,

420 F(1,84)=13.037, p= 0.0005) and their contributions varied across sources (ANOVA,

421 F(3,84)=237.618, p<0.0001) but not across floral organs (ANOVA, F(2,84)=1.758, p=0.1787; Fig.

422 5; Table S3). Comparisons between the focal floral organ and the remaining flower organs had

423 the highest values of nestedness and lowest values of turnover than comparisons with all other

424 potential sources. Soil, instead had the lowest values of nestedness, and the highest turnover (Fig.

425 5). This analysis suggests that the OTUs in our focal communities are (to some extent) a subset

426 of those present in other flowers in the co-flowering community, and have a number of OTUs

427 not present (or present in low abundance) in our soil samples.

428 Consistent with these results, using a Bayesian approach to track the potential sources of

429 microbes (Knights et al. 2011), a large proportion of our sample was assigned to be from the co-

430 flowering community (neighbors) and only a small percentage from soil and leaves. All groups

431 (organ and treatment) had a large proportion of bacterial taxa that was assigned to unknown

432 sources, with a larger contribution in the petals.

19 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

433

434 Discussion

435 Despite their importance for plant community function and fitness, we know very little about the

436 communities of microbes that inhabit flowers. Here we showed that different organs within a

437 flower have different epiphytic bacterial communities which overwhelm the effects of

438 geographic distance on community composition. Our results indicate that bacterial communities

439 of flowers are established by the balance of dispersal and environmental selection, but this

440 balance will be different for each organ within a flower. We suggest that floral organs (especially

441 the petals) act as environmental filters and that, in the absence of pollinators, the metacommunity

442 as a whole might be better described within a species-sorting paradigm that emphasizes niche

443 differences (Leibold et al. 2004). However, our data suggest that, within organ environmental

444 selection could become overwhelmed by pollinator-mediated dispersal of new taxa (especially in

445 organs with extensive pollinator engagement like the stamens) and, with high rates of visitation

446 the metacommunity might be better described through a mass-effect perspective, were

447 metacommunity dynamics are mostly determined by dispersal (Leibold et al. 2004).

448

449 Environmental selection

450 Consistent with previous studies of floral microbiomes, we found that flowers of the yellow

451 monkeyflower (M. guttatus) have microbial communities that are distinct from other plant

452 organs (Fig. S6; Junker et al. 2011; Ottesen et al. 2013; Junker & Keller 2015; Wei & Ashman

453 2018). We provide evidence that floral organs act as different environmental habitats

454 contributing to the assembly of flower microbiomes, despite the small size of M. guttatus

455 flowers, and the close contact between stamens and styles with the petals. These results confirm

20 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

456 predictions based on knowledge of chemical and morphological differences of these floral parts

457 (Aleklett et al. 2014; Junker & Keller 2015). Specifically, 1) floral organ explains more of the

458 variation in community assembly than seep or pollination treatment, 2) OTUs within a particular

459 flower organ are more phylogenetically clustered than expected by random, and 3) most

460 differences in community composition do not correlate with distance at the scale of this study.

461 Moreover, a neutral model fails to explain the patterns of distribution of a large proportion of

462 OTUs within the flower. Thus, our results corroborate previous work showing that floral organs

463 support different microbial communities (Junker & Keller 2015) and that flower

464 microenvironments (i.e., nectar) can act as strong environmental filters (Herrera et al. 2010) but

465 also extend them by separating the effect of different organs and measuring the effect of

466 dispersal on the effectiveness of organ selection.

467 We found that a large proportion of OTUs were shared among two or more organs (more

468 than expected by chance for OTUs that deviate from a neutral expectation) and that differences

469 in epiphytic bacterial community composition across organs could be accounted by the

470 nestedeness component of beta diversity. These observations suggest that each organ

471 microbiome is a subset of monkeyflower metacommunity and that, potentially, each organ filter

472 acts in a (more or less) sequential manner. Within the flower, we hypothesize that petals act as

473 the first environmental filter. Petal microbial communities had the highest proportion of unique

474 taxa and showed the strongest signals of selection (i.e., they had a larger proportion of over and

475 underrepresented taxa; Fig. 3B). While some bacteria taxa might be enriched in the flowers, were

476 they are able to grow on floral volatiles and other carbon compounds (Abanda-Nkpwatt et al.,

477 2006), it is likely that the strongest selection is to get rid of potential pathogens and other

478 microbes with potentially negative effects on the plant fitness. Monkeyflower petals have a

21 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

479 much larger area than stigmas or stamens and are exposed to a larger proportion of microbes

480 coming from neighboring flowers, transferred by bees, or moved passively from the soil and

481 other organs of the plant. However, once on the petals, microbes could be filtered by petal traits

482 (like pigments, volatiles, trichomes and epidermal cell shapes) that can affect antibiotic

483 properties, surface water retention and temperature of the flower (Whitney et al. 2011; Harrap et

484 al. 2017; Cisowska et al. 2011; Huang et al. 2011). From the petals microbes could colonize the

485 style and stamens. In the case of the style at least, the presence of pollinators increases the rates

486 of microbe colonization from the petals: style communities in the control (open to pollinators)

487 treatment tend to be more similar to petal communities and share more OTUs than those in the

488 pollination exclusion treatment (Fig. 2A; Fig. 2C).

489

490 Dispersal

491 While it is often assumed that pollinators play a key role in dispersal of microbial communities,

492 and some systems bear this out (e.g. yeast nectar communities; Pozo et al. 2014; Vannette &

493 Fukami 2017), insects visiting flowers have diverse interactions with organs within a flower

494 (Plowright & Laverty 1984). In this study, pollinator treatment on its own did not account for

495 differences in community composition, but rather affected the importance of organ selection in

496 explaining differences in community composition across samples. A similar study looking at

497 community composition of microbiomes of whole tomato flowers found no differences between

498 pollinator exclusion and their control treatments but found increased variation across flower

499 microbiomes in the control (pollinators allowed) relative to flowers in the absence of pollinators

500 (Allard et al. 2018). Here we show that the magnitude of the effect depends on the floral organ as

501 well as the rate and type of interaction with pollinators.

22 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

502 The interplay of dispersal and environmental selection has been hard to disentangle in the

503 field (Evans et al. 2017). Here, we showed that differences in visitation rate of insects to yellow

504 monkeyflower explained some of the variation in treatment differences among locations for the

505 stamen samples. However, the intimacy of the interaction also played a role. Some of these

506 visitors were butterflies and flies that rarely contacted internal organs of the flowers (Fig. 4B),

507 while others had more extensive internal contact with the floral organs. Indeed, the bacterial

508 communities of stamens and styles showed the largest differences between pollination treatments

509 and, in particular, the bacterial communities in the styles had a larger proportion of unique OTUs

510 in the exclusion treatment relative to the control.

511 Mimulus guttatus flowers in the field are mostly visited by medium and large bees

512 foraging for pollen (Robertson et al. 1999; Wu et al. 2008; Arceo-Gómez et al. 2018). Thus,

513 pollinators are likely to have sustained engagement with the stamens and alter the microbial

514 environment of these organs by removing pollen. In a recent paper, Russell et al. (2019) showed

515 that in flowers of M. guttaus scrabbling (one of the common behaviors to forage pollen in bees)

516 results in a larger deposition of microbes than other behaviors, and in artificial flowers this

517 behavior leads to the largest deposition of bacteria on the stamen. Here, we showed that

518 differences in bacterial community composition in the stamens across pollinator treatments

519 increased with increased pollination rates. This correlation was clearer when we considered

520 abundance and phylogenetic distance in our analysis. These results are consistent with our

521 observation that, in the absence of pollinators, phylogenetic diversity of the bacterial

522 communities of stamens is much lower than in the presence of pollinators. Finally, the bacterial

523 communities of the stamens in the presence of pollinators are more consistent with a neutral

23 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

524 pattern than in our exclusion treatment, and consistent with a contribution of local dispersal,

525 bacterial communities in the anthers became more differentiated with increased distance.

526 Overall, in this system, petals seem to be acting as a major environmental filter where

527 only a few bacterial taxa can establish and, while a few new bacteria might colonize at high rates

528 of visitation, many of these taxa will remain in low abundance, unable to establish as part of the

529 main petal community. Instead, pollinator engagement with the stamens can introduce variation

530 in the communities and outweigh some of the contributions of environmental selection.

531 One caveat however, is that amplicon studies can underestimate the effect of organ

532 selection because it is not possible to distinguish dormant species, which can represent a large

533 proportion of microbial communities (Jones and Lennon 2010; Lennon and Jones 2011).

534 Dormancy can facilitate dispersal (Locey 2010) and minimize the experienced environmental

535 stressors (Jones and Lennon 2010) potentially obscuring signals of organ level selection. With

536 the exception of some nectar yeasts and bacteria, and some floral pathogens, we do not know the

537 extent to which microbes are actively growing in flowers. Of the orders we observed in high

538 abundance, many (e.g., Bacillales, Clostridiales, Actinomycetales) are characterized by taxa with

539 spores or other forms of dormancy (e.g., Paredes-Sabja et al. 2011). Future studies should

540 address the proportion of dormant cells in different organs of the flower, the relative contribution

541 of pollinators to that dormant pool as well as the functional roles of microbes in different parts of

542 the flower.

543 Furthermore, we did not see a signal of community differentiation by distance (except in

544 the stamens of the control treatment), and while one might be tempted to conclude that microbes

545 are “everywhere”, it might instead reflect limited resolution of 16S sequences within the spatial

546 scales chosen in this study. The small sizes of microbes could mean different spatial scales at

24 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

547 which environmental selection and dispersal shape local communities. On the one hand, small

548 cells mean that even within a single organ within the plant, microbes might be experiencing

549 many different environments (Lindow & Brandl 2003). The strongest signal of niche sorting

550 might occur at the sub-organ scales. On the other hand, because of their small size dispersal of

551 some microbes might occur at much larger scales than the ones considered in this project

552 (Wilkinson et al. 2012; Choudoir et al. 2018).

553 Differentiation by distance alone is not the best indicator of neutrality because different

554 patterns of dispersal will result in different patterns of spatial differentiation of communities (e.g.

555 pollinators might travel only a few meters or more than a kilometer in a single dispersal event;

556 Castilla et al. 2017) and locations separated by distance might also experience slightly different

557 environments. This might also be why we did not observe a relationship between the co-

558 flowering community at each site and the microbial communities in the flowers of M. guttatus in

559 those same sites. In this study we tried to minimize these effects by having the small and medium

560 length distances replicated along different environments and directions (Fig. 1A, B) and by using

561 different lines of evidence to asses neutrality (see Environmental Selection section).

562 While this study provides explicit measurements of neutral and selective contributions of

563 microbial communities of flowers in the presence and absence of pollinators, it also highlights

564 that most of the variation in community composition of floral microbiomes remains unexplained.

565 Factors outside the flower (including soil chemistry) could affect the local pool of microbes or

566 even the floral chemistry (Majetic et al. 2008; Meindl et al. 2014). Similarly, variation across

567 flowers, across plants, or even within a single plant due to competition and strong priority effects

568 (e.g. Peay et al. 2012) could be contributing to the unexplained effects and unfortunately, much

25 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

569 of that variation is obscured in this study because to obtain enough DNA we had to pool together

570 the organs of three different flowers from the same cage for each sample.

571 Here we have shown the importance of “intimacy” and rate of pollination for microbial

572 dispersal in different organs. However, consistent with recent results (Allard et al. 2018), we

573 have also shown that floral communities have a similar composition in the presence or absence

574 of pollinators indicating the importance of other mechanisms of microbial dispersal in shaping

575 floral colonization (e.g., wind, soil and rain). These unknown sources (wind, water, florivores,

576 nectar robbers, other nearby flowers) all could have contributed to the large proportion of OTUs

577 that we were unable to assign to a known source.

578 Unfortunately, another source of unknown microbes that can play a role (especially for

579 small samples like some of the ones used in this study) is contamination during sampling,

580 extraction and sequencing. While is possible that we had some contamination (it is common in

581 low biomass microbiome analyses; Eisenhofer et al. 2018) we were unable to amplify any of our

582 controls, and to minimize the effects of minor contaminants, we randomized our samples and

583 used sterile equipment at every stage of the process. Similarly, while contamination could

584 explain the presence of some human associated taxa, it could also due to imperfect taxonomy

585 assignment that depends what is already in the database. Additionally, it could be that the species

586 present (which were not able to identify) can be found in flowers. While most of what we know

587 of Streptoccocus comes from human pathogens, this genus has also been found in the aerial

588 surfaces of plants (e.g. Pontonino et al., 2018).

589 Finally, the best practices in analyzing microbiome data are still subject of much debate

590 (e.g. Callahan et al. 2017; Pollock et al. 2018). In this study we analyzed the data in multiple

591 ways (standardizing libraries vs. rarefying data; using OTUs vs. error learning ASV assignment;

26 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

592 see methods) and in most cases the method did not affect the results, and in the cases in which it

593 often provided different levels of resolution and different information (same when varying

594 diversity indices). In some cases, these multiple analyses provided added confidence in the

595 results (e.g. community differences between organs). Whereas in other cases (e.g. the

596 relationships between visitation rates and pairwise beta diversity in petal and style samples)

597 discrepancies between analyses suggest caution is warranted.

598 This study advances our understanding of community assembly of flower microbiomes. It

599 highlights the interplay between dispersal and environmental selection, providing insight into

600 potential effects of pollinator disturbance or floral changes on microbial community

601 composition. As the reproductive structure of angiosperms, microbial effects on flowers can

602 have a large impact on plant fitness. From previous studies we know that microbes of flowers

603 can modify volatile production and nectar composition affecting pollinator visitation (e.g.

604 Herrera et al. 2008; Rering et al. 2017). In addition, the flower is the main site of infection of

605 important pathogens of plants (e.g. anther smut, Erwinia amylovora) and microbial communities

606 of flowers can affect the probability of infection (e.g. Pseudomonas fluorescens growth on

607 significantly reduces the establishment of Erwinia; Wilson and Lindow, 1992).

608 Despite its importance, the flower remains a relatively understudied environment for

609 microbes and there is still much we do not know. Future studies should address the effects of

610 different floral traits and floral heterogeneity in the assembly of microbial communities, the

611 importance of these microbes to plant fitness and the effects of microbial community assembly

612 on plant communities and the evolution of plant traits. A better understanding of the processes

613 affecting community assembly of flower-associated microbiomes provides insight into the

27 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

614 processes driving flower-microbe-pollinator interactions and the potential effects of different

615 disturbances and environmental changes in changing these dynamics.

616

617

618 Aknowledgments

619 We are grateful with R. Hayes for field assistance and with P. Aigner and C. Koehler for the

620 facilities provided at the McLaughlin Natural Reserve. Instructors at the EDAMAME workshop,

621 A. Johnson and C. Marshall all provided helpful advice for the analyses. Members of the

622 Ashman lab provided useful feedback at different parts during this project. Feedback from two

623 anonymous reviewers greatly improved this manuscript. This work was supported by the

624 National Science Foundation (DEB 1452386 to T-L. Ashman). M. Rebolleda-Gómez was

625 supported by the Dietrich School of Arts and Sciences through a Pittsburgh Ecology and

626 Evolution Postdoctoral fellowship.

627

28 bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

References

Abanda-Nkpwatt D., Müsch M., Tschiersch J., Boettner M., & Schwab W. (2006). Molecular interaction 589 between Methylobacterium extorquens and seedlings: growth promotion, methanol consumption, 590 and localization of the methanol emission site. J Exp Bot., 57, 4025–4032.

Aleklett, K., Hart, M., & Shade, A. (2014). The microbial ecology of flowers: an emerging frontier in phyllosphere research. Botany, 92(4), 253–266.

Alexander, H. M., & Antonovics, J. (1995). Spread of anther-smut disease (Ustilago-Violacea) and character correlations in a genetically variable experimental population of Silene Alba. J. Ecol., 83(5), 783–794.

Allard, S. M., Ottesen, A. R., Brown, E. W., & Micallef, S. A. (2018). Insect exclusion limits variation in bacterial microbiomes of tomato flowers and fruit. J. App. Microbiol., Doi: 10.111/jam.14087.

Anders, S., & Huber, W. (2010). Differential expression analysis for sequence count data. Genome Biol., 11(10), R106. DOI: 10.1186/gb-2010-11-10-r106

Anderson, M. J. (2006). Distance-based tests for homogeneity of multivariate dispersions. Biometrics, 62(1), 245–253.

Arceo-Gómez, G., & Ashman, T. L. (2011). Heterospecific pollen deposition: Does diversity alter the consequences? New Phytol., 192(3), 738–746

Arceo-Gómez, G., Kaczorowski R. L., & Ashman, T. L. (2018). A network approach to understanding patterns of coflowering in diverse communities. International Journal of Plant Sciences, 179(7), 569-582.

Arceo-Gómez, G., & Ashman, T.-L. (2014). Coflowering Community Context Influences Female Fitness and Alters the Adaptive Value of Flower Longevity in Mimulus guttatus. Am. Nat., 183(2), E50–E63.

Baselga, A. (2010). Partitioning the turnover and nestedness components of beta diversity. Global Ecol. and Biogeogr., 19(1), 134–143.

Baselga, A., & Orme, C. D. L. (2012). Betapart: An R package for the study of beta diversity. Methods Ecol. and Evol., 3(5), 808–812.

Belisle, M., Peay, K. G., & Fukami, T. (2012). Flowers as Islands: Spatial Distribution of Nectar- Inhabiting Microfungi among Plants of Mimulus aurantiacus, a Hummingbird-Pollinated Shrub. Microb. Ecol., 63(4), 711–718.

* bioRxiv preprint doi: https://doi.org/10.1101/721647; this version posted August 1, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license.

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A prectical and powerful approach to multiple testing. J. R. Stat. Soc., 57, 289–300.

Bokulich, N. A., Subramanian, S., Faith, J. J., Gevers, D., Gordon, J. I., Knight, R., Mills, D. A., et al. (2012). Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. Nat. Methods, 10(1), 57-59.

Burns, A. R., Stephens, W. Z., Stagaman, K., Wong, S., Rawls, J. F., Guillemin, K., & Bohannan, B. J. M. (2016). Contribution of neutral processes to the assembly of gut microbial communities in the zebrafish over host development. ISME J, 10, 655–664.

Burns, A. R., Miller, E., Agarwal, M., Rolig, A. S., Milligan-Myhre, K., Seredick, S., et al. (2017). Interhost dispersal alters microbiome assembly and can overwhelm host innate immunity in an experimental zebrafish model. PNAS, 114(42), 11181-11186.

Callahan, B.J., McMurdie, P. J., Rosen M. J., Han, A.W., Johnson, A. J. A., Holmes, S. P. (2016). DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods, 13, 581-583.

Callahan, B. J., McMurdie, P. J., Holmes, S.P. (2017). Exact sequence variants should replace operational taxonomic units in marker-gene data analysis. ISME J, 11, 2639-2643.

Caporaso, J.G., Bittinger, K., Bushman,F.D., DeSantis, T. Z., Andersen, G. L. and R. Kinght. (2010). PyNAST: a flexible tool for aligning sequences to a template alignment. Nat. Methods, 7, 335–336.

Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman,F.D., Costello, E.K. et al. (2010b). QIIME allows analysis of high-throughput community sequencing data. Nat. Methods, 7, 335–336.

Caporaso, J.G. Lauber, C. L., Walters, W. A., Berg-Lyons, D. Huntley, J. Fierer, N. et al. (2012). Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 6, 1621-1624.

Chappell, C. R. & Fukami, T. (2018). Nectar yeasts: A natural microcosm for ecology. Yeast. Doi: 10.1002/yea.3311.

Chase, J. M. & Leibold, M. A. (2003). Ecological niches: linking classical and contemporary approaches. The University of Chicago Press, Chicago, IL.

Chen, H., & Boutros, P. C. (2011). VennDiagram: A package for the generation of highly- customizable Venn and Euler diagrams in R. BMC Bioinf., 12(1), 35. DOI: 10.1186/1471- 2105-12-35.

Chesson, P. (2000). Mechanisms of Maintenance of Species Diversity. Annu. Rev. Ecol. Evol. Syst., 31, 343–366.

Choudoir, M. J., Barberán, A., Menninger, H. L., Dunn, R. R., & Fierer, N. (2018). Variation in range size and dispersal capabilities of microbial taxa. Ecology, 99(2), 322–334.

Cisowska, A., Wojnicz, D., & Hendrich, A. B. (2011). Anthocyanins as antimicrobial agents of natural plant origin. Nat. Prod. Commun., 6(1), 149–156.

Costello, E. K., Stagaman, K., Dethlefsen, L., Bohannan, B. J. M., & Relman, D. A. (2012). The application of ecological theory towards an understanding of the human microbiome. Science, 336(6086), 1255–1262.

Donaldson, G. P., Ladinsky, M. S., Yu, K. B., Sanders, J. G., Yoo, B. B., Chou, W. C., et al. (2018). Gut microbiota utilize immunoglobulin a for mucosal colonization. Science, 360(6390), 795–800.

Economo, E. P., & Keitt, T. H. (2008). Species diversity in neutral metacommunities: A network approach. Ecol. Lett., 11(1), 52–62.

Eisenhofer, R., Minich, J.J., Marotz, C., Cooper, A, Kight R., & Weyrich, L. S. (2018). Contamination in low microbial biomass microbiome studies: issues and recommendations. Trends in Microbiology, 27, 105-117.

Evans, S., Martiny, J. B. H., & Allison, S. D. (2017). Effects of dispersal and selection on stochastic assembly in microbial communities. ISME J., 11(1), 176–185.

Faith D. P. (1992). Conservation evaluation and phylogenetic diversity. Biol. Conserv. 61, 1–10.

Fitzpatrick, C., Lu-Irving, P., Copeland, J., Guttman, D. S., Wang, P. W., Baltrus, D. A., et al. (2018). Chloroplast sequence variation and the efficacy of peptide nucleic acids for blocking host amplification in plant microbiome studies. Microbiome, 6, 144. DOI: 10.1186/s40168- 018-0534-0.

Freestone, A. L., and B. D. Inouye. (2006). Dispersal limitation and environmental heterogeneity shape scale-dependent diversity patterns in plant communities. Ecology, 87, 2425–2432.

Harrap, M. J. M., Rands, S. A., de Ibarra, N. H., & Whitney, H. M. (2017). The diversity of floral temperature patterns, and their use by pollinators. ELife, 6, e31262. DOI: 10.7554/eLife.31262.

Harrison, S., J. Maron, & G. Huxel. (2000). Regional turnover and fluctuation in populations of five plants confined to serpentine seeps. Conserv. Biol. 14,769–779.

Herrera, C. M., Canto, A., Pozo, M. I., & Bazaga, P. (2010). Inhospitable sweetness: Nectar filtering of pollinator-borne inocula leads to impoverished, phylogenetically clustered yeast communities. Proc. R. Soc. B, 277(1682), 747–754.

Herrera, C. M., Pozo, M. I., & Medrano, M. (2013). Yeasts in nectar of an early-blooming herb: Sought by bumble bees, detrimental to plant fecundity. Ecology, 94, 273–279.

Hubbell, S.P. (2001). The Unified Neutral Theory of Biodiversity and Biogeography. Princeton University Press, Princeton, NJ.

Huang, M., Sanchez-Moreiras, A. M., Abel, C., Sohrabi, R., Lee, S., Gershenzon, J., & Tholl, D. (2012). The major volatile organic compound emitted from Arabidopsis thaliana flowers, the sesquiterpene (E)-β-caryophyllene, is a defense against a bacterial pathogen. New Phytol., 193(4), 997–1008.

Jones, S. E., & Lennon, J. T. (2010). Dormancy contributes to the maintenance of microbial diversity. PNAS, 107(13), 5881–5886.

Junker, R. R., Loewel, C., Gross, R., Dötterl, S., Keller, A., & Blüthgen, N. (2011). Composition of epiphytic bacterial communities differs on petals and leaves. Plant Biology, 13(6), 918– 924.

Junker, R. R., & Keller, A. (2015). Microhabitat heterogeneity across leaves and flower organs promotes bacterial diversity. FEMS Microbiol. Ecol., 91(9), fiv097. Doi: 10.1093/femsec/fiv097

Kembel, S.W., Cowan, P.D., Helmus, M.R., Cornwell, W.K., Morlon, H., Ackerly, D.D., et al. (2010). Picante: R tools for integrating phylogenies and ecology. Bioinformatics. 26, 1463- 1464.

Knights, D., Kuczynski, J., Charlson, E. S., Zaneveld, J., Mozer, M. C., Collman, R. G., et al. (2011). Bayesian community-wide culture-independent microbial source tracking. Nature Methods. 8, 761-763.

Koski, M. H., Meindl, G. A., Arceo-Gómez, G., Wolowski, M., LeCroy, K. A., & Ashman, T. L. (2015). Plant–flower visitor networks in a serpentine metacommunity: assessing traits associated with keystone plant species. Arthropod-Plant Interactions, 9(1), 9–21.

Laverty T. M., Plowright R. C. (1988). Flower handling by bumblebees: a comparison of specialists and generalists. Anim Behav, 36, 733–740.

Leibold, M. A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J. M., Hoopes, M. F., et al. (2004). The metacommunity concept: A framework for multi-scale community ecology. Ecol. Letters, 7, 601–613.

Leibold, M. A., & McPeek, M. A. (2006). Coexistence of the niche and neutral perspectives in community ecology. Ecology, 87(6), 1399–1410.

Lennon, J. T., & Jones, S. E. (2011). Microbial seed banks: The ecological and evolutionary implications of dormancy. Nat. Rev. Microbiol. 9, 119–130.

Lindow, S. E., & Brandl, M. T. (2003). Microbiology of the phyllosphere. Appl. Environ. Microbiol., 69(4), 1875–1883.

Locey, K. J. (2010). Synthesizing traditional biogeography with microbial ecology: The importance of dormancy. J. Biogeogr., 37(10), 1835–1841.

Love, M. I., Huber, W., & Anders, S. (2014). Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol, 15(12), 1–21. DOI: 10.1186/s13059-014-0550-8

Lowe, W. H. & McPeek, M. A. (2014). Is dispersal neutral? TREE, 29(8), 444-450.

Lozupone, C., & Knight, R. (2005). UniFrac: a new phylogenetic method for comparing microbial communities. Appl. Environ. Microbiol. 71, 8228–8235.

Lundberg DS, Yourstone S, Mieczkowski P, Jones CD, Dangl JL. (2013). Practical innovations for high-throughput amplicon sequencing. Nat. Methods, 10(10), 999–1002.

MacArthur, R. H., & Wilson, E. O. (1967). The theory of island biogeography. Princeton University Press, Princeton, N.J

Majetic, C. J., Raguso, R. A., & Ashman, T. L. (2008). The impact of biochemistry vs. population membership on floral scent profiles in colour polymorphic Hesperis matronalis. Ann. Bot., 102(6), 911–922.

Meindl, G. A., Bain, D. J., & Ashman, T. L. (2014). Nickel accumulation in leaves, floral organs and rewards varies by serpentine soil affinity. AoB PLANTS, 6(Brooks 1987), 1–9. https://doi.org/10.1093/aobpla/plu036

Mittelbach, M., Yurkov, A. M., Stoll, R., & Begerow, D. (2016). Inoculation order of nectar- borne yeasts opens a door for transient species and changes nectar rewarded to pollinators. Fungal Ecol., 22, 90–97.

Morrison-Whittle, P., & Goddard, M. R. (2015). Quantifying the relative roles of selective and neutral processes in defining eukaryotic microbial communities. ISME J, 9(9), 2003–2011.

Nemergut, D. R., Schmidt, S. K., Fukami, T., O’Neill, S. P., Bilinski, T. M., Stanish, L. F., … Ferrenberg, S. (2013). Patterns and Processes of Microbial Community Assembly. Microbiol. Mol. Biol. Rev., 77(3), 342–356.

O’Dwyer, J. P., Kembel, S. W., & Green, J. L. (2012). Phylogenetic Diversity Theory Sheds Light on the Structure of Microbial Communities. PLoS Comput. Biol., 8(12). DOI: 10.1371/journal.pcbi.1002832.

Ofek-Lalzar, M., Sela, N., Goldman-Voronov, M., Green, S. J., Hadar, Y., & Minz, D. (2014). Niche and host-associated functional signatures of the root surface microbiome. Nat. Commun., 5, 1–9. DOI:doi.org/10.1038/ncomms5950

Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D., et al. (2018). vegan: Community Ecology Package. R package version 2.4-6. https://CRAN.R- project.org/package=vegan

Ottesen, A. R., Peña, A. G., White, J. R., Pettengill, J. B., Li, C., Allard, S., et al. (2013). Baseline survey of the anatomical microbial ecology of an important food plant: Solanum lycopersicum (tomato). BMC Microbiol., 13, 114.

Paredes-Sabja, D., Setlow, P., & Sarker, M. R. (2011). Germination of spores of Bacillales and Clostridiales species: Mechanisms and proteins involved. Trends Microbiol., 19(2), 85–94.

Peay K.G., Belisle, M., & Fukami, T. (2012). Phylogenetic relatedness predicts priority effects in nectar yeast communities. Proc Biol Sci., 22, 749-758.

Plowright, R. C., & Laverty, T. M. (1984). The ecology and sociobiology of bumble bees. Annu. Rev. Ento., 29, 175–199.

Pollock, J., Glendinning L., Wisedchanwet, T., & Watson M. (2018). The madness of microbiome: attempting to find consensus for “best practice” for 16s microbiome studies. Appl Environ Microbiol. 84: e02627-17.

Pontonino, E., Di Cagno R., Tarraf W., Filannino P., De Mastro G., & Gobbetti M. (2018). Dynamic and assembly of epiphyte and endophyte lactic acid bacteria during the life cycle of Origanum vulgare L. Front. Microbiol., 9: 1372.

Pozo, M. I., Herrera, C. M., & Alonso, C. (2014). Spatial and temporal distribution patterns of nectar-inhabiting yeasts: How different floral microenvironments arise in winter-blooming Helleborus foetidus. Fungal Ecol., 11, 173–180.

Pozo, M. I., Herrera, C. M., Lachance, M. A., Verstrepen, K., Lievens, B., & Jacquemyn, H. (2016). Species coexistence in simple microbial communities: unravelling the phenotypic landscape of co-occurring Metschnikowia species in floral nectar. Environ. Microbiol. 18, 1850–1862.

Pratte, Z. A., Besson, M., Hollman, R. D., & Stewarta, F. J. (2018). The gills of reef fish support a distinct microbiome influenced by hostspecific factors. Appl. Environ. Microbiol., 84(9). DOI: 10.1128/AEM.00063-18.

R Core Team. (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Rawls, J. F., Mahowald, M. A., Ley, R. E., & Gordon, J. I. (2006). Reciprocal Gut Microbiota Transplants from Zebrafish and Mice to Germ-free Recipients Reveal Host Habitat Selection. Cell, 127(2), 423–433.

Rering, C. C., Beck, J. J., Hall, G. W., McCartney, M. M., & Vannette, R. L. (2017). Nectar- inhabiting microorganisms influence nectar volatile composition and attractiveness to a generalist pollinator. New Phytol. 220, 655–658.

Russell, A., Rebolleda-Gómez, T. M. Shaible and T-L. Ashman. 2019. Movers and shakers: bumble bee foraging behavior shapes the dispersal of microbes among and within flowers. Ecosphere, 10, e02714

Robertson, A. W., Mountjoy, C., Faulkner, B. E., Roberts, M. V., & Macnair, M. R. (1999). Bumble bee selection of Mimulus guttatus flowers: The effects of pollen quality and reward depletion. Ecology, 80(8), 2594–2606.

Schaeffer, R. N., Mei, Y. Z., Andicoechea, J., Manson, J. S., & Irwin, R. E. (2017). Consequences of a nectar yeast for pollinator preference and performance. Funct. Ecol., 31(3), 613–621.

Schaeffer, R. N., Phillips, C. R., Duryea, M. C., Andicoechea, J., & Irwin, R. E. (2014). Nectar yeasts in the tall Larkspur Delphinium Barbeyi (Ranunculaceae) and effects on components of pollinator foraging behavior. PLoS ONE, 9(10), e108214.

Schluter, J., & Foster, K. R. (2012). The Evolution of Mutualism in Gut Microbiota Via Host Epithelial Selection. PLoS Biol., 10(11), e1001424.

Shade, A., McManus, P. S., & Handelsman, J. (2013). Unexpected Diversity during Community Succession in the Apple. MBio, 4, e00602-12. DOI:doi.org/10.1128/mBio.00602

Sloan, W. T., Lunn, M., Woodcock, S., Head, I. M., Nee, S., & Curtis, T. P. (2006). Quantifying the roles of immigration and chance in shaping prokaryote community structure. Enviro. Microbiol., 8(4), 732–740.

Sloan, W. T., Woodcock, S., Lunn, M., Head, I. M., & Curtis, T. P. (2007). Modeling taxa- abundance distributions in microbial communities using environmental sequence data. Microb. Ecol., 53(3), 443–455.

Soininen, J., McDonald, R., & Hillebrand, H. (2007). The distance decay of similarity in ecological communities. Ecography, 30(1), 3–12.

Tilman, D. (2004). Niche tradeoffs, neutrality, and community structure: a stochastic theory of resource competition, invasion, and community assembly. PNAS, 101(30), 10854–10861.

Vannette, R. L., Gauthier, M.-P. L., & Fukami, T. (2013). Nectar bacteria, but not yeast, weaken a plant-pollinator mutualism. Proc. R. Soc. B, 280(1752), 20122601–20122601.

Vannette, R. L., & Fukami, T. (2017). Dispersal enhances beta diversity in nectar microbes. Ecol. Letters. 20, 901-910.

Vellend, M. (2016). The Theory of Ecological Communities. Princeton University Press, Princeton, NJ.

Venkataraman, A., Bassis, C. M., Beck, J. M., Young, V. B., Curtis, J. L., Huffngale, G. B., Schmidt, T. M. (2015). Application of a neutral community model to asses structurig of the human lung microbiome. mBio. 6(1), e02284-14. DOI: 10.1128/mBio.02284-14

Wei, N., & Ashman, T. L. (2018). The effects of host species and sexual dimorphism differ among root, leaf and flower microbiomes of wild strawberries in situ. Sci. Rep., 8(1), 1–12. DOI: 10.1038/s41598-018-23518-9

Wilson, M., & Lindow, S. E. (1992). Interactions between the biological control agent Pseudomonas fluorescens A506 and Erwinia amylovora in pear blossoms. Phytopathology, 83:117-123.

Whitney, H. M., Bennett, K. M. V., Dorling, M., Sandbach, L., Prince, D., Chittka, L., & Glover, B. J. (2011). Why do so many petals have conical epidermal cells? Ann. Bot, 108(4), 609– 616.

Wilkinson, D. M., Koumoutsaris, S., Mitchell, E. A. D., & Bey, I. (2012). Modelling the effect of size on the aerial dispersal of microorganisms. J. Biogeogr., 39(1), 89–97.

Wu, C. A., Lowry, D. B., Cooley, A. M., Wright, K. M., Lee, Y. W., & Willis, J. H. (2008). Mimulus is an emerging model system for the integration of ecological and genomic studies. Heredity, 100, 220–230.

Zhang, J., Kobert, K., Flouri, T. & Stamatakis, A. (2014). PEAR: a fast and accurate Illumina Paired-End read merger. Bioinformatics. 30, 614-620.

Data availability: DNA sequences will be made available through NCBI’s Short Read Archive,

all other data will be deposited on Dryad. R code for analyses is available on GitHub at

github.com/mrebolleda/OrganFilters_MimulusMicrobiome

Authorship: MRG and TLA designed the research. MRG conducted observations, sampling and

analyses. MRG wrote the manuscript with substantial input from TLA.

* certified bypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmadeavailableunder bioRxiv preprint

Table 1. Table with PERMANOVA summaries using different betadiversity indices (Org- Floral organ; Pol- Pollinator doi:

Sorensen Bray-Curtis Unifrac Weighted Unifrac https://doi.org/10.1101/721647

Df F R2 p F R2 p F R2 p F R2 p

Floral organ (Org) 2 3.364 0.068 0.001 1.993 0.042 0.001 5.543 0.107 0.001 1.913 0.039 0.016

Pollinator (Pol) 1 1.273 0.013 0.143 1.085 0.011 0.289 1.299 0.013 0.142 1.313 0.014 0.206 ; Seep 4 1.326 0.054 0.020 1.157 0.049 0.113 1.396 0.054 0.024 1.532 0.063 0.022 a this versionpostedAugust1,2019. CC-BY-ND 4.0Internationallicense

Org*Pol 2 1.451 0.03 0.033 1.359 0.029 0.040 1.557 0.03 0.019 1.63 0.034 0.038 . treatment). The copyrightholderforthispreprint(whichwasnot

* certified bypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmadeavailableunder bioRxiv preprint

Org*Seep 8 1.03 0.084 0.347 1.122 0.095 0.113 1.161 0.09 0.077 1.128 0.093 0.204

Pol*Seep 4 0.96 0.039 0.614 0.886 0.037 0.800 0.846 0.033 0.832 0.837 0.035 0.722 doi:

Residuals 70 0.712 0.737 0.675 0.722 https://doi.org/10.1101/721647

; a

this versionpostedAugust1,2019. CC-BY-ND 4.0Internationallicense . The copyrightholderforthispreprint(whichwasnot

Figure 1. Experimental design. A. Map of seeps at McLaughlin Natural Reserve and sampling locations within them. The different colors indicate each seep. B. Diagram of sampling locations within each seep. Distances between sampling location as 10-100m. C. Picture of pollinator exclusion cage and control cage in one of the locations. D. Diagram of Mimulus gutattus flower showing the floral organs studied: petals (pink), the four stamens (yellow) and the style (turquoise). A. Seep Elevation 0 500 1000 m BNS 0 - 325 TP9 325 - 488 RH1 488 - 651 RHA > 651 RHB

Lake

Yolo

Napa

TP9 RH1 RHA RHB BNS

C. D.

Figure 2. Contribution of floral organ differences to bacterial community structure of monkeyflowers (Mimulus guttatus). A. PCoA based on Unifrac differences between samples. Each point represents one sample. B. Bacterial communities of each organ are phylogenetically clustered (i.e., lower phylogenetic diversity than expected). C. Venn diagrams of the core microbiome of each floral organ for each of the pollinator treatments. D. Alpha diversity across floral organs. Small circles represent each sample and the large circles the mean values. A. B. 2 Pollinator treatment 0.2 Exclusion Control

0.1 0 Floral organ Petals Stamens 0.0 Style -2 PCoA2 (8%) PCoA2 -0.1

-0.2 -4 Phylogenetic diversity (estandarized effect) (estandarized diversity Phylogenetic -0.2 0.0 0.2 PCoA1(17%) C. D.

3 Exclusion Control Style Style 3 7 13 13 7 2 34 Petals 5 27 741 Petals 3 36 22 5 1 Stamens Stamens (Shannon) diversity Alpha

Petals Stamens Style

Figure 3. Effect of pollinator treatment in beta diversity across floral organs of Mimulus guttatus. A. Beta diversity calculated in a multivariate space from Bray-Curtis distances. B. Relationship between geographical distance and betadiversity (Bray-Curtis) across different floral structures and leaves. C. Proportion of OTUs that are shared across at least two floral organs. The distribution is the result of 1000 simulations assuming no differences between groups other than their sizes. The vertical dashed line indicates the observed values. Control A. Pollinator exclusion 1.0

0.8

0.6

0.4 Beta diversity (Bray-Curtis) diversity Beta

Stamens Petals Style Leaf 012012012012 Distance (km) B. 100 Style Overrepresented Stamens Petals Underrepresented e 2 2 2 g R =0.37 R =0.20 R =0.34 Neutral a t n

e 75 c r e p 50 e c n e r r

u 25 c O 0 -12 -8 -4 -12 -8 -4 -12 -8 -4 log(mean relative abundance)

C. Overrepresented Underrepresented Neutral 1.00

0.50 Frequency

(null distribution) (null 0.00 0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 Proportion of OTUs shared across at least two floral organs

Figure 4. Pollinator effect on community composition. A. Percentage variation in community composition explained by organ in the control and pollinator exclusion treatments. B. Composition of pollinator pool (percentage of total number of visits) contacting external or internal surfaces of the flower (Functional pollinator categories from Koski et al., 2015: XS- extra-large social bees; LS- Large social bees; SB- small solitary bee; MB- medium solitary bee; LB- large solitary bee (pollen on body); LL- large solitary bee (pollen on legs); BF- bee fly; SF- small syrphid fly; LF- large syrphid fly; FL- other flies; LE- moths and butterflies; BE- beetles; VE- wasps; OT- other). C. Correlations between visitation rate to the plots and community composition differences between pollinator treatments at each location (pairwise beta diversity- see small insert). Colors show the strength of the correlation and the numbers inside denote the p- values after adjustment for multiple testing (see text for details). Analyses were performed using different beta diversity indices. Sørensen (Sor) and Unifrac (Uni) include only presence-absence data whereas Bray-Curtis (B-C) and weighted Unifrac (W-Uni) also account for relative abundances. A. B. Exclusion 100 XS Control LS SB MB 15 75 LB LL 12 50 BF SF 9 LF Percentage FL 25 LE 6 BE VE 0 OT SorB-C Uni W-Uni

% Variation explained by "organ" External Internal Beta diversity index

C. All visits Internal r 0.8 0.6 0.56 0.65 0.65 0.001 0.26 0.26 0.4 0.2 0.65 0.65 0.72 0.05 0.06 0.51 0.0 Abundance B-C W-Uni

0.65 0.65 0.21 0.18 0.77 0.04 Uni Beta diversity index diversity Beta

Presence 0.65 0.65 0.65 0.22 0.18 0.18 Sor

Pairwise beta diversity beta Pairwise 01020 Stamens Style Petals Stamens Style Petals Visitation rate (visits per plot/hr)

Figure 5. Potential sources of microbial communities in M. guttatus flowers. A-C. Components of beta diversity comparing potential source microbial communities (rest of the M. guttatus flower, M. guttatus leaves, floral neighbors, soil) against the communities of stamens, petals and styles (in columns). A. Total beta diversity as measured by the Sørensen index. B. Portion of beta diversity that is due to turnover of species. C. Component of beta diversity that is due to nestedness. Points are the mean ± one standard deviation. Solid circles are samples from the control treatment and solid circles are from the pollinator exclusion treatment. D. Proportion of OTUs in each treatment that could be assigned to each source using Bayesian estimation. The rest of the flower was not added as a potential source in this analysis.